ADHD Adults Diagnosis

ADHD Adults Diagnosis

library(dplyr)

Attaching package: 'dplyr'

The following objects are masked from 'package:stats':

    filter, lag

The following objects are masked from 'package:base':

    intersect, setdiff, setequal, union

library(ellipse)

Warning: package 'ellipse' was built under R version 4.5.2

Attaching package: 'ellipse'

The following object is masked from 'package:graphics':

    pairs

library(ggplot2)

Warning: package 'ggplot2' was built under R version 4.5.2

library(ggrepel)

Warning: package 'ggrepel' was built under R version 4.5.2

library(kableExtra)

Attaching package: 'kableExtra'

The following object is masked from 'package:dplyr':

    group_rows

library(metafor)

Loading required package: Matrix

Warning: package 'Matrix' was built under R version 4.5.2

Loading required package: metadat

Loading required package: numDeriv

Loading the 'metafor' package (version 4.8-0). For an
introduction to the package please type: help(metafor)

library(stringr)

Warning: package 'stringr' was built under R version 4.5.2

library(tidyr)

Warning: package 'tidyr' was built under R version 4.5.2

Attaching package: 'tidyr'

The following objects are masked from 'package:Matrix':

    expand, pack, unpack

library(writexl)

Warning: package 'writexl' was built under R version 4.5.2

data_location <- "https://github.com/jeremymiles/adhd_git/raw/refs/heads/main/data/diagnosis_data.csv"

d <- read.csv(data_location)

### TEMP - REMOVE
#d <- d %>%
#  dplyr::filter(ID != "Harrison, 2019{#351}") 

capitalize_first <- function(s) {
  if (is.na(s) || s == "") { #Handle NAs and empty strings
    return(s)
  }
  first_letter <- str_sub(s, 1, 1)
  rest_of_string <- str_sub(s, 2)
  return(str_c(str_to_upper(first_letter), rest_of_string))
}

You can add options to executable code like this

#|fig-height: 12
d_ss <- d %>% 
  dplyr::select(
    ID, size, starts_with("sensitivity"), starts_with("specificity"),
    test_description_self
  ) %>%
  dplyr::select(
    ID, size,
    ends_with(
      c("self_report", "neuropsycho_tests", "clinician_interview",
        "combination", "clinician_tool", "clinician_rating", 
        "peer_rating", "neuroimaging", "neuroimaging", 
        "observational", "biomarker", "EEG")
    )
  ) %>%
  dplyr::select(
    !starts_with(c(
      "sensitivity_CI", "specificity_CI", 
      "sensitivity_other", "specificity_other"))
  )

d_test_description_self <- d %>%
  dplyr::select(
    ID, test_description_self
  ) %>% 
  dplyr::distinct()

d_sens <- d_ss %>%
  dplyr::select(ID, starts_with("sens")
  )

d_spec <- d_ss %>%
  dplyr::select(ID, starts_with("spec")
  )

d_sens_long <- d_sens %>%
  tidyr::pivot_longer(
    cols = -ID
  ) %>%
  dplyr::rename(sensitivity = value)%>%
  dplyr::mutate(name = stringr::str_remove(name, "sensitivity_"))



d_spec_long <- d_spec %>%
  tidyr::pivot_longer(
    cols = -ID
  ) %>%
  dplyr::rename(specificity = value) %>%
  dplyr::mutate(name = stringr::str_remove(name, "specificity_"))


d_ss_long <- dplyr::full_join(
  d_sens_long, d_spec_long
) %>% 
  dplyr::filter(
    !is.na(sensitivity) & !is.na(specificity)
  ) %>% 
  dplyr::mutate(
    name = sapply(name, capitalize_first)
  ) %>%
  dplyr::mutate(
    name = stringr::str_replace_all(name, "_", " "),
    name = ifelse(
      name == "Neuropsycho tests", "Neuropsychological Test", name
    ),
    name = ifelse(
      name == "Self report", "Self-Report Questionnaire", name
    ),
    name = ifelse(
      name == "Peer rating", "Peer-Rating", name
    ),
    name = ifelse(
      name == "Clinician interview", "Clinician Tool", name
    ),
    
  )

Joining with `by = join_by(ID, name)`

# mark neurotypicals
d_neurotypical <- d %>% dplyr::select(ID, Neurotypical, size) %>% 
  dplyr::mutate(
    Neurotypical = Neurotypical == "Neurotypical"
  )

d_neurotypical_Only <- d %>% 
  dplyr::mutate(
    Neurotypical_Only = Neurotypical == "Neurotypical" &
      Clinical != "Clinical" &
      Autism != "Autism" &
      Antisocial != "Antisocial" &
      Depression != "Depression" &
      Feigning != "Feigning"
  ) %>% dplyr::select(ID, Neurotypical_Only, size)


# count the number of different types of study
d_name_count <- data.frame(dplyr::bind_rows(
  c(name = "Combination", count = sum(d$Combination == "Combination")),
  c(name = "Biomarker", count = sum(d$Biomarker == "Biomarker")),
  c(name = "Clinician Tool", count = sum(d$Clinician.interview == "Clinician interview")),
  c(name = "EEG", count = sum(d$EEG == "EEG")),
  c(name = "Neuroimaging", count = sum(d$Neuroimaging == "Neuroimaging")),
  c(name = "Neuropsychological Test",   count = sum(d$Neuropsychological == "Neuropsychological")),
  c(name = "Peer-Rating",   count = sum(d$Peer.report == "Peer report")),
  c(name = "Self-Report Questionnaire",   count = sum(d$Self.report == "Self report"))
))

# merge with long data
d_ss_long <- d_ss_long %>%
  dplyr::full_join(
    d_neurotypical
  ) %>% 
  dplyr::full_join(
    d_neurotypical_Only
  ) %>% 
  dplyr::filter(!is.na(name))

Joining with `by = join_by(ID)`
Joining with `by = join_by(ID, size)`

ggplot2::ggplot(
  d_ss_long, 
  aes(
    x = sensitivity, 
    y = specificity)) +
  geom_point() +
  xlab("Sensitivity (%)") + 
  ylab("Specificity (%)") +
  xlim(c(0, 100)) + 
  ylim(c(0, 100))

d_ss_long <- d_ss_long %>%
  dplyr::full_join(d_name_count, by = "name") %>%
  dplyr::mutate(
    name_with_count = paste0(name, " (n = ", count, " studies)")
  )

ggplot2::ggplot(
  d_ss_long, 
  aes(
    x = sensitivity, 
    y = specificity,
    colour = name, shape = Neurotypical)) +
  geom_point(size = 2.5) +
  xlab("Sensitivity (%)") + 
  ylab("Specificity (%)") +
  facet_wrap(~ name_with_count) +
  guides(colour = "none") +
  theme(legend.position = "bottom") +
  xlim(c(0, 100)) + 
  ylim(c(0, 100))

#Figure 8
ggplot2::ggplot(
  d_ss_long, 
  aes(
    x = sensitivity, 
    y = specificity,
    colour = name, shape = Neurotypical_Only)) +
  geom_point(size = 2.5) +
  xlab("Sensitivity") + 
  ylab("Specificity") +
  facet_wrap(~ name, nrow = 2) +
  guides(colour = "none") +
  labs(
    shape = "Neurotypical only"
  ) +
  xlim(c(0, 100)) + 
  ylim(c(0, 100))

# separate charts
for (name_1 in d_ss_long$name) {
  plot <- ggplot2::ggplot(
    d_ss_long %>% dplyr::filter(name_1 == name), 
    aes(
      x = sensitivity, 
      y = specificity,
      shape = Neurotypical_Only)) +
    geom_point(size = 2.5) +
    xlab("Sensitivity") + 
    ylab("Specificity") +
    guides(colour = "none") +
    labs(
      shape = "Neurotypical only"
    ) +
    ggtitle(name_1) +
    xlim(c(0, 100)) + 
    ylim(c(0, 100))
  print(plot)
}

# Figure 8 separate charts

Self report only

# this is figure 5
cat("self report<p>")

self report<p>

# 1. Prepare your data first
plot_data <- d_ss_long %>%
  dplyr::filter(
    name == "Self-Report Questionnaire"
  ) %>%
  dplyr::inner_join(
    d_test_description_self
  ) %>%
  dplyr::mutate(
    test_description = word(test_description_self, 1),
    size = size)

Joining with `by = join_by(ID)`

# Set colors for some
plot_data <- plot_data %>%
  dplyr::mutate(
    `Frequent Tools` = "Other",
    `Frequent Tools` = ifelse(substr(test_description, 1, 5) == "CAARS",
                              "CAARS", `Frequent Tools`),
    `Frequent Tools` = ifelse(substr(test_description, 1, 5) == "BAARS", 
                              "BAARS", `Frequent Tools`),
    `Frequent Tools` = ifelse(substr(test_description, 1, 4) == "ASRS", 
                              "ASRS", `Frequent Tools`),
    `Frequent Tools` = ifelse(substr(test_description, 1, 4) == "WURS", 
                              "WURS", `Frequent Tools`)
  )

plot_data %>%
  dplyr::group_by(test_description, `Frequent Tools`) %>%
  dplyr::summarise(n = dplyr::n()) %>%
  dplyr::arrange(desc(n)) %>%
  knitr::kable()

`summarise()` has grouped output by 'test_description'. You can override using
the `.groups` argument.

test_description	Frequent Tools	n
CAARS-S:L	CAARS	5
ASRS-A	ASRS	4
BAARS-IV	BAARS	3
ASRS-v1.1	ASRS	2
CAARS-S	CAARS	2
ADHD-items	Other	1
ADSA	Other	1
AHA	Other	1
ALS-SF	Other	1
APQ	Other	1
ASRS	ASRS	1
ASRS-5	ASRS	1
ASSET-BS	Other	1
BADDS	Other	1
CAARS-AI	CAARS	1
CAARS-S:SV	CAARS	1
CBS	Other	1
EarlyDetect(ASRS)	Other	1
IPDE-SQ-11	Other	1
Online	Other	1
PAI	Other	1
PDI-4	Other	1
SR-WRAADDS	Other	1
WURS	WURS	1
WURS-25	WURS	1
WURS-4	WURS	1
WURS-61	WURS	1
WURS/CAARS	WURS	1

# 1. Define your colors in a named vector
# You must assign a color to every category, or the others will disappear/turn grey.
my_colors <- c(
  "ASRS"  = "red",    # Replace with your preferred color
  "CAARS" = "blue",   # Replace with your preferred color
  "BAARS" = "darkgreen",  # Replace with your preferred color
  "WURS" = "brown4",  # Replace with your preferred color
  "Other" = "black"   # The specific requirement
)

# 2. Plot
ggplot2::ggplot(
  data = plot_data, 
  aes(
    x = sensitivity,
    y = specificity,
    shape = Neurotypical_Only,
    size = size,
    color = `Frequent Tools`
  )
) +
  geom_point() +
  # 3. Add the manual scale
  scale_color_manual(values = my_colors) +
  xlab("Sensitivity (%)") +
  ylab("Specificity (%)") +
  ggrepel::geom_text_repel(
    aes(label = test_description), max.overlaps = 10, size = 3) +
  labs(
    shape = "Neurotypical only"
  ) +
  scale_size_continuous(
    name = "Size", # This sets the legend title
    breaks = c(50, 500, 1000), # Original untransformed values for legend
    labels = c("50", "500", "1000"), # Corresponding labels
    guide = "legend", # Explicitly request a legend
    # Set limits based on the *transformed* size values from your plot_data
    # This is crucial for the scale to match your actual plotted points
  ) + 
  theme(legend.position = "right") +
  xlim(c(0, 100)) + 
  ylim(c(0, 100))

d_test_description_np <- d %>%
  dplyr::select(
    ID, test_description_neuropsycho_tests, size
  ) %>% 
  dplyr::distinct() %>% 
  dplyr::mutate(
    test_description = word(test_description_neuropsycho_tests, 1)
  ) 

d_test_description_np %>%
  dplyr::group_by(test_description) %>%
  dplyr::summarise(n = dplyr::n()) %>%
  dplyr::arrange(desc(n)) %>%
  knitr::kable()

test_description	n
	99
Model	8
QbTest	4
AQT	2
Battery	2
C-CPT-II	2
Go/NoGo	2
MOXO-dCPT	2
BQSS	1
C-CPT	1
C-CPT-3	1
DS	1
IVA+Plus-FSRCQ	1
Measures	1
QbTest-Plus	1
SCWT	1
Stroop	1
TOAD	1
TOVA	1

rm(plot_data)

# This is figure 6
cat("Neuropsychological \n Tests<p>")

Neuropsychological 
 Tests<p>

plot_data <- d_ss_long %>%
  dplyr::filter(
    name == "Neuropsychological Test"
  ) %>%dplyr::inner_join(
    d_test_description_np
  ) %>%
  dplyr::mutate(test_description = word(test_description_neuropsycho_tests, 1)) 

Joining with `by = join_by(ID, size)`

# Set colors for some
plot_data <- plot_data %>%
  dplyr::mutate(
    `Frequent Tools` = "Other",
    `Frequent Tools` = 
      ifelse(substr(test_description, 1, 3) == "AQT", "AQT", `Frequent Tools`),
    `Frequent Tools` = 
      ifelse(substr(test_description, 1, 5) == "C-CPT", "C-CPT", `Frequent Tools`),
    `Frequent Tools` = 
      ifelse(substr(test_description, 1, 4) == "MOXO", "MOXO", `Frequent Tools`),
    `Frequent Tools` = 
      ifelse(substr(test_description, 1, 6) == "QbTest", "QbTest", `Frequent Tools`
      ),
    `Frequent Tools` = 
      ifelse(substr(test_description, 1, 5) == "Go-No", "Go-NoGo", `Frequent Tools`)
  )

my_colors <- c(
  "AQT"  = "red",    # Replace with your preferred color
  "C-CPT" = "blue",   # Replace with your preferred color
  "MOXO" = "darkgreen",  # Replace with your preferred color
  "QbTest" = "chocolate4",
  "Go-NoGo" = "yellow",
  "Other" = "black"   # The specific requirement
)

plot_data %>% ggplot2::ggplot(
  aes(
    x = sensitivity, 
    y = specificity,
    shape = Neurotypical_Only, size = size,
    color = `Frequent Tools`)
) +
  geom_point() +
  scale_color_manual(values = my_colors) +
  xlab("Sensitivity (%)") + 
  ylab("Specificity (%)") +
  geom_text_repel(aes(label = test_description), max.overlaps = 10, size = 3) +
  labs(
    shape = "Neurotypical only"
  ) + 
  theme(legend.position = "right") +
  xlim(c(0, 100)) + 
  ylim(c(0, 100))

AUC_vars <- 
  d %>% dplyr::select(
    starts_with("AUC") 
  ) %>%
  dplyr::select(-starts_with("AUC_CI")) %>%
  names()
AUC_vars

 [1] "AUC_self_report"         "AUC_peer_rating"        
 [3] "AUC_neuropsycho_tests"   "AUC_neuroimaging"       
 [5] "AUC_clinician_interview" "AUC_feigningADHD"       
 [7] "AUC_observational"       "AUC_combination"        
 [9] "AUC_DMV"                 "AUC_clinician_tool"     
[11] "AUC_biomarker"           "AUC_general_cognitive"  
[13] "AUC_EEG"                

d <- d %>%
  dplyr::mutate(
    group = dplyr::case_when(
      Neurotypical == "Neurotypical" &
        Clinical != "Clinical" &
        Autism != "Autism" &
        Antisocial != "Antisocial" &
        Depression != "Depression" &
        Feigning != "Feigning" ~ "Neurotypical",
      (Clinical == "Clinical" |
         Autism == "Autism" |
         Antisocial == "Antisocial" |
         Depression == "Depression") &
        Neurotypical != "Neurotypical" &
        Feigning != "Feigning" ~ "Clinical",
      Neurotypical == "Neurotypical" & (
        Clinical == "Clinical" |
          Autism == "Autism" |
          Antisocial == "Antisocial" |
          Depression == "Depression" |
          Feigning == "Feigning" # Include Feigning here if needed for "Both"
      ) ~ "Both (or more)",
      TRUE ~ NA_character_
    )
  ) 



d_long_accuracy <- d %>%
  dplyr::select(
    accuracy_biomarker,
    accuracy_combination,
    accuracy_EEG,
    accuracy_neuroimaging,
    accuracy_neuropsycho_tests,
    accuracy_observational,
    accuracy_peer_rating,
    accuracy_self_report,
    group
  ) %>% 
  dplyr::filter(!is.na(group)) %>%
  pivot_longer(
    cols = starts_with("accuracy_"),
    names_to = "accuracy_type",
    values_to = "accuracy_value",
    names_prefix = "accuracy_"
  )%>% 
  dplyr::mutate(measure = "accuracy") %>%
  dplyr::rename_with(~ gsub("accuracy_", "", .x), starts_with("accuracy"))




d_long_AUC <- d %>%
  dplyr::select(
    starts_with("AUC"),
    group
  ) %>% 
  dplyr::mutate(
    AUC_neuropsycho_tests = as.numeric(AUC_neuropsycho_tests)
  ) %>% 
  dplyr::select(-starts_with("AUC_CI")) %>% 
  dplyr::filter(!is.na(group)) %>%
  pivot_longer(
    cols = starts_with("AUC_"),
    names_to = "AUC_type",
    values_to = "AUC_value",
    names_prefix = "AUC_"
  ) %>% 
  dplyr::mutate(
    AUC_value = ifelse(
      AUC_value < 1, AUC_value * 100, AUC_value
    )
  ) %>% 
  dplyr::mutate(measure = "AUC") %>%
  dplyr::rename_with(~ gsub("AUC_", "", .x), starts_with("AUC")) %>%     
  dplyr::filter(!is.na(value)) %>% 
  dplyr::filter(type != "feigningADHD")

d_long_both <- dplyr::bind_rows(
  d_long_accuracy, d_long_AUC
) %>% dplyr::filter(!is.na(value))

# Make labels nicer.

d_long_both <- d_long_both %>%
  dplyr::mutate(
    type = dplyr::case_when(
      type == "biomarker" ~ "Biomarker",
      type == "clinician_interview" ~ "Clinician\nTool",
      type == "combination" ~ "Combination",
      type == "neuroimaging" ~ "Neuroimaging",
      type == "observational" ~ "Observational",
      type == "neuropsycho_tests" ~ "Neuro-\npsychological",
      type == "peer_rating" ~ "Peer\nRating",
      type == "self_report" ~ "Self-Report",
      .default = type
    ),
    measure = ifelse(measure == "accuracy", "Accuracy", measure)
  )



d_long_both %>% dplyr::filter(!is.na(value)) %>%
  dplyr::rename(Group = group) %>%
  ggplot2::ggplot(
    aes(x = Group, y = value)
  )  + 
  geom_boxplot() +
  facet_grid(type ~ measure)+
  theme(axis.text.x = element_text(angle = 90, hjust = 1))

# Figure 7
d_long_both %>% dplyr::filter(!is.na(value)) %>%
  dplyr::rename(Group = group) %>%
  ggplot2::ggplot(
    aes(x = Group, y = value)
  )  + 
  geom_boxplot() +
  facet_grid(measure ~ type)+
  theme(axis.text.x = element_text(angle = 90, hjust = 1))+
  labs(y = NULL)

d_long_both %>% dplyr::filter(!is.na(value)) %>%
  dplyr::rename(Group = group) %>%
  ggplot2::ggplot(
    aes(x = Group, y = value, color = measure)
  )  + 
  geom_boxplot() +
  facet_wrap(~ type) +
  theme(axis.text.x = element_text(angle = 90, hjust = 1))

Function to get Accuracy

# Function to read CSV and select specific columns
read_adhd_data <- function(file_path) {
  
  # Read the CSV file
  data <- read.csv(file_path)
  
  # Select only the required columns
  clean_data <- data %>%
    select(n_ADHD, 
           sensitivity_self_report, 
           specificity_self_report,
           size, ID)
  
  return(clean_data)
}

########################### Usage example
accuracy <- read_adhd_data(data_location)
#############################

# Check the result
head(accuracy)

  n_ADHD sensitivity_self_report specificity_self_report size
1    518                      NA                      NA  845
2     38                      NA                      NA   69
3    142                      92                      NA  280
4   1006                      NA                      NA 1135
5     63                      89                      13   69
6     70                      81                      71  140
                    ID
1   Abramson, 2023{#5}
2     Adamou, 2022{#8}
3      Aita, 2018{#16}
4      Amen, 2021{#28}
5    Bakare, 2020{#45}
6 Bastiaens, 2017{#55}

calculate_adhd_stats <- function(data) {
  
  processed_data <- data %>%
    mutate(
      # 1. Calculate Prevalence (P / Total)
      Prevalence = n_ADHD / size,
      
      # 2. Calculate (1 - Prevalence)
      One_Minus_Prevalence = 1 - Prevalence,
      
      # 3. Create cleaner columns for Sensitivity/Specificity for the formula
      Sensitivity = sensitivity_self_report,
      Specificity = specificity_self_report,
      
      # 4. Calculate Accuracy
      # Formula: (Sens * Prev) + (Spec * (1 - Prev))
      Accuracy = (Sensitivity * Prevalence) + (Specificity * One_Minus_Prevalence)
    )
  
  return(processed_data)
}
accuracy <- calculate_adhd_stats(accuracy)
head(accuracy)

  n_ADHD sensitivity_self_report specificity_self_report size
1    518                      NA                      NA  845
2     38                      NA                      NA   69
3    142                      92                      NA  280
4   1006                      NA                      NA 1135
5     63                      89                      13   69
6     70                      81                      71  140
                    ID Prevalence One_Minus_Prevalence Sensitivity Specificity
1   Abramson, 2023{#5}  0.6130178           0.38698225          NA          NA
2     Adamou, 2022{#8}  0.5507246           0.44927536          NA          NA
3      Aita, 2018{#16}  0.5071429           0.49285714          92          NA
4      Amen, 2021{#28}  0.8863436           0.11365639          NA          NA
5    Bakare, 2020{#45}  0.9130435           0.08695652          89          13
6 Bastiaens, 2017{#55}  0.5000000           0.50000000          81          71
  Accuracy
1       NA
2       NA
3       NA
4       NA
5  82.3913
6  76.0000

accuracy <- accuracy %>%
  select(-Sensitivity, -Specificity)

# Create and save in one pipe chain
output_file <- "accuracy.html"


# Create and save
kableExtra::kbl(accuracy) %>%
  kable_styling(bootstrap_options = c("striped", "hover")) %>%
  save_kable(file = output_file)

Program Summary of Findings Table

# Test: combination
# Outcome: Clinical_misdiagnosis



CreateProgSummary <- function(test_outcome_and_result) {
  
  if (test_outcome_and_result == "cost_self_report") {
    test_outcome_and_result <- "cost_self.report"
  }
  
  reverse_min_max <- 
    stringr::str_detect(test_outcome_and_result, "sensitivity") | 
    stringr::str_detect(test_outcome_and_result, "specificity") 
  
  print(test_outcome_and_result)
  
  if (test_outcome_and_result %in% 
      c(
        "admin_self_report", 
        "admin_neuropsycho_tests", 
        "cost_neuropsycho_tests",
        "concordance_self_report",
        "concordance_neuropsycho_tests",
        "kappa_neuropsycho_tests",
        "kappa_neuroimaging")
  ) {
    return(
      data.frame(
        test_outcome_and_result = test_outcome_and_result,
        contributing_studies = "Misnamed variable (probably doesn't matter?)",
        primary_results  = "Misnamed variable"
      )
    )
  }
  
  
  non_numeric_question <- !is.numeric(d[[test_outcome_and_result]]) 
  
  d_now <-
    d %>%
    dplyr::mutate(!!rlang::sym(test_outcome_and_result) := as.numeric(!!rlang::sym(test_outcome_and_result))) %>%
    dplyr::filter(!is.na(!!rlang::sym(test_outcome_and_result)))
  
  n <- nrow(d_now)
  min_value <- min(d_now[[test_outcome_and_result]])
  max_value <- max(d_now[[test_outcome_and_result]])
  
  min_id <- d_now[which(d_now[[test_outcome_and_result]] == min_value), "Refid"] %>% paste0(., collapse = ", ")
  max_id <- d_now[which(d_now[[test_outcome_and_result]] == max_value), "Refid"] %>% paste0(., collapse = ", ")
  
  
  contributing_studies <- (glue::glue(
    "{paste(d_now$ID)}")
  ) %>% as.character() %>% paste(., collapse = "; ")
  
  if(non_numeric_question) {
    contributing_studies <- primary_results <- "Warning: non-numeric result."
  } else {
    contributing_studies <- 
      glue::glue("{n} studies ({contributing_studies})")
    if (!reverse_min_max) {
      primary_results <- 
        glue::glue(
          "{min_value}(#{min_id}) to {max_value}(#{max_id})"
        )
    } else {
      primary_results <- 
        glue::glue(
          "{max_value}(#{max_id}) to {min_value}(#{min_id})"
        )
    }
    
  }
  
  
  return(
    data.frame(
      test_outcome_and_result = test_outcome_and_result,
      contributing_studies = contributing_studies,
      primary_results = primary_results))
}


kq_outcome <- c(
  "clinical_misdiagnosis",
  "sensitivity",
  "specificity",
  "admin",
  "kappa",
  "ICC",
  "cost",
  "concordance"
)

kq_index_test <- c(
  "combination",
  "self_report",
  "peer_rating",
  "neuropsycho_tests",
  "neuroimaging",
  "EEG",
  "biomarker",
  "observational",
  "clinician_interview",
  "clinician_tool",
  "feigningADHD"
)  

all_kqs <- 
  expand_grid(kq_index_test, kq_outcome ) %>% 
  dplyr::select(kq_outcome, kq_index_test) %>% 
  apply(., 1, paste0, collapse = "_")



res <- lapply(all_kqs, CreateProgSummary)

[1] "clinical_misdiagnosis_combination"
[1] "sensitivity_combination"
[1] "specificity_combination"
[1] "admin_combination"

Warning: There was 1 warning in `dplyr::mutate()`.
ℹ In argument: `admin_combination = as.numeric(admin_combination)`.
Caused by warning:
! NAs introduced by coercion

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf

Warning in max(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "kappa_combination"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "ICC_combination"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "cost_combination"

Warning: There was 1 warning in `dplyr::mutate()`.
ℹ In argument: `cost_combination = as.numeric(cost_combination)`.
Caused by warning:
! NAs introduced by coercion

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf

Warning in max(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "concordance_combination"

Warning: There was 1 warning in `dplyr::mutate()`.
ℹ In argument: `concordance_combination = as.numeric(concordance_combination)`.
Caused by warning:
! NAs introduced by coercion

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf

Warning in max(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "clinical_misdiagnosis_self_report"
[1] "sensitivity_self_report"
[1] "specificity_self_report"
[1] "admin_self_report"
[1] "kappa_self_report"
[1] "ICC_self_report"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "cost_self.report"

Warning: There was 1 warning in `dplyr::mutate()`.
ℹ In argument: `cost_self.report = as.numeric(cost_self.report)`.
Caused by warning:
! NAs introduced by coercion

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf

Warning in max(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "concordance_self_report"
[1] "clinical_misdiagnosis_peer_rating"
[1] "sensitivity_peer_rating"
[1] "specificity_peer_rating"
[1] "admin_peer_rating"

Warning: There was 1 warning in `dplyr::mutate()`.
ℹ In argument: `admin_peer_rating = as.numeric(admin_peer_rating)`.
Caused by warning:
! NAs introduced by coercion

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf

Warning in max(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "kappa_peer_rating"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "ICC_peer_rating"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "cost_peer_rating"

Warning: There was 1 warning in `dplyr::mutate()`.
ℹ In argument: `cost_peer_rating = as.numeric(cost_peer_rating)`.
Caused by warning:
! NAs introduced by coercion

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf

Warning in max(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "concordance_peer_rating"

Warning: There was 1 warning in `dplyr::mutate()`.
ℹ In argument: `concordance_peer_rating = as.numeric(concordance_peer_rating)`.
Caused by warning:
! NAs introduced by coercion

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf

Warning in max(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "clinical_misdiagnosis_neuropsycho_tests"
[1] "sensitivity_neuropsycho_tests"
[1] "specificity_neuropsycho_tests"
[1] "admin_neuropsycho_tests"
[1] "kappa_neuropsycho_tests"
[1] "ICC_neuropsycho_tests"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "cost_neuropsycho_tests"
[1] "concordance_neuropsycho_tests"
[1] "clinical_misdiagnosis_neuroimaging"
[1] "sensitivity_neuroimaging"
[1] "specificity_neuroimaging"
[1] "admin_neuroimaging"

Warning: There was 1 warning in `dplyr::mutate()`.
ℹ In argument: `admin_neuroimaging = as.numeric(admin_neuroimaging)`.
Caused by warning:
! NAs introduced by coercion

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf

Warning in max(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "kappa_neuroimaging"
[1] "ICC_neuroimaging"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "cost_neuroimaging"

Warning: There was 1 warning in `dplyr::mutate()`.
ℹ In argument: `cost_neuroimaging = as.numeric(cost_neuroimaging)`.
Caused by warning:
! NAs introduced by coercion

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf

Warning in max(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "concordance_neuroimaging"

Warning: There was 1 warning in `dplyr::mutate()`.
ℹ In argument: `concordance_neuroimaging =
  as.numeric(concordance_neuroimaging)`.
Caused by warning:
! NAs introduced by coercion

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf

Warning in max(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "clinical_misdiagnosis_EEG"
[1] "sensitivity_EEG"
[1] "specificity_EEG"
[1] "admin_EEG"

Warning: There was 1 warning in `dplyr::mutate()`.
ℹ In argument: `admin_EEG = as.numeric(admin_EEG)`.
Caused by warning:
! NAs introduced by coercion

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf

Warning in max(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "kappa_EEG"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "ICC_EEG"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "cost_EEG"

Warning: There was 1 warning in `dplyr::mutate()`.
ℹ In argument: `cost_EEG = as.numeric(cost_EEG)`.
Caused by warning:
! NAs introduced by coercion

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf

Warning in max(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "concordance_EEG"

Warning: There was 1 warning in `dplyr::mutate()`.
ℹ In argument: `concordance_EEG = as.numeric(concordance_EEG)`.
Caused by warning:
! NAs introduced by coercion

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf

Warning in max(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "clinical_misdiagnosis_biomarker"
[1] "sensitivity_biomarker"
[1] "specificity_biomarker"
[1] "admin_biomarker"

Warning: There was 1 warning in `dplyr::mutate()`.
ℹ In argument: `admin_biomarker = as.numeric(admin_biomarker)`.
Caused by warning:
! NAs introduced by coercion

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf

Warning in max(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "kappa_biomarker"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "ICC_biomarker"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "cost_biomarker"

Warning: There was 1 warning in `dplyr::mutate()`.
ℹ In argument: `cost_biomarker = as.numeric(cost_biomarker)`.
Caused by warning:
! NAs introduced by coercion

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf

Warning in max(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "concordance_biomarker"

Warning: There was 1 warning in `dplyr::mutate()`.
ℹ In argument: `concordance_biomarker = as.numeric(concordance_biomarker)`.
Caused by warning:
! NAs introduced by coercion

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf

Warning in max(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "clinical_misdiagnosis_observational"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "sensitivity_observational"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "specificity_observational"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "admin_observational"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "kappa_observational"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "ICC_observational"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "cost_observational"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "concordance_observational"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "clinical_misdiagnosis_clinician_interview"
[1] "sensitivity_clinician_interview"
[1] "specificity_clinician_interview"
[1] "admin_clinician_interview"

Warning: There was 1 warning in `dplyr::mutate()`.
ℹ In argument: `admin_clinician_interview =
  as.numeric(admin_clinician_interview)`.
Caused by warning:
! NAs introduced by coercion

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf

Warning in max(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "kappa_clinician_interview"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "ICC_clinician_interview"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "cost_clinician_interview"

Warning: There was 1 warning in `dplyr::mutate()`.
ℹ In argument: `cost_clinician_interview =
  as.numeric(cost_clinician_interview)`.
Caused by warning:
! NAs introduced by coercion

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf

Warning in max(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "concordance_clinician_interview"

Warning: There was 1 warning in `dplyr::mutate()`.
ℹ In argument: `concordance_clinician_interview =
  as.numeric(concordance_clinician_interview)`.
Caused by warning:
! NAs introduced by coercion

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf

Warning in max(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "clinical_misdiagnosis_clinician_tool"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "sensitivity_clinician_tool"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "specificity_clinician_tool"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "admin_clinician_tool"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "kappa_clinician_tool"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "ICC_clinician_tool"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "cost_clinician_tool"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "concordance_clinician_tool"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "clinical_misdiagnosis_feigningADHD"
[1] "sensitivity_feigningADHD"
[1] "specificity_feigningADHD"
[1] "admin_feigningADHD"

Warning: There was 1 warning in `dplyr::mutate()`.
ℹ In argument: `admin_feigningADHD = as.numeric(admin_feigningADHD)`.
Caused by warning:
! NAs introduced by coercion

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf

Warning in max(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "kappa_feigningADHD"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "ICC_feigningADHD"

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf
Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "cost_feigningADHD"

Warning: There was 1 warning in `dplyr::mutate()`.
ℹ In argument: `cost_feigningADHD = as.numeric(cost_feigningADHD)`.
Caused by warning:
! NAs introduced by coercion

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf

Warning in max(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

[1] "concordance_feigningADHD"

Warning: There was 1 warning in `dplyr::mutate()`.
ℹ In argument: `concordance_feigningADHD =
  as.numeric(concordance_feigningADHD)`.
Caused by warning:
! NAs introduced by coercion

Warning in min(d_now[[test_outcome_and_result]]): no non-missing arguments to
min; returning Inf

Warning in max(d_now[[test_outcome_and_result]]): no non-missing arguments to
max; returning -Inf

writexl::write_xlsx(dplyr::bind_rows(res), "adhd_adults_diagnosis_SoF.xlsx")

Sensitivity and Specificity Meta-Analysis

#
d$test_description_self_report <- word(d$test_description_self, 1)
d$test_description_peer_rating <- d$test_description_peer


# If you want to clean up any trailing spaces left behind:
d <- d %>%
  dplyr::mutate(
    study = gsub("\\{.*\\}", "", ID),
    study = trimws(study),
    study = ifelse(study == "van de Glind, 2013", "Van de Glind, 2013", study),
    test_description_self_report = test_description_self
  )

Need to loop through:

self_report peer_rating neuropsycho_tests neuroimaging clinician_interview combination biomarker EEG

#| fig-width: 14
#| fig-height: 14

inv_logit <- function(x) { exp(x) / (1 + exp(x)) }
logit <- function(x) { log(x / (1 - x)) }


plot_diagnostic_meta <- 
  function(
    data, diag_type_name, 
    true_pos, false_neg, 
    true_neg, false_pos,
    instrument_col
  ) {
    
    data[[instrument_col]] <- stringr::word(data[[instrument_col]])
    
    dat_sens <- metafor::escalc(
      measure = "PLO", 
      xi = data[[true_pos]],                       # Access the column via name
      ni = (data[[true_pos]] + data[[false_neg]]), # Calculate the denominator
      data = data, 
      add = 0.5
    )
    dat_sens$outcome <- "sens"
    
    # Calculate Specificity
    dat_spec <- metafor::escalc(
      measure = "PLO", 
      xi = data[[true_neg]], 
      ni = (data[[true_neg]] + data[[false_pos]]), 
      data = data, 
      add = 0.5
    )
    dat_spec$outcome <- "spec"
    
    # Combine and filter for studies that actually have data for this type
    dat_long <- rbind(dat_sens, dat_spec)
    dat_long <- dat_long %>% 
      filter(!is.na(yi)) %>%
      arrange(desc(study))
    
    if(nrow(dat_long) == 0) return(message(paste("No data for", diag_type_name)))
    
    # 2. Run Meta-Analysis Model
    res <- metafor::rma.mv(yi, vi, mods = ~ outcome - 1, random = ~ outcome | study,
                           struct = "UN", data = dat_long)
    
    # 3. Setup Plotting Layout
    all_studies <- dat_long %>%
      distinct(study) %>%
      # Add the instrument label early so we can sort by it
      mutate(instrument_label = data[[instrument_col]][match(study, data$study)]) %>%
      # SORTING: Instrument (A-Z) and then Study (A-Z)
      # We arrange in ASCENDING order here...
      arrange(instrument_label, study) %>%
      # ...then we assign row_num so that the first item gets the HIGHEST row 
      # (putting it at the top of the plot)
      mutate(row_num = rev(row_number()))
    
    k_total <- nrow(all_studies)
    cex <- 0.6
    
    layout(matrix(c(1, 2), nrow = 1), widths = c(1.8, 1))
    
    # 4. Plot Sensitivity
    par(mar = c(5, 4, 3, 1))
    dat_sens_plot <- 
      all_studies %>% 
      left_join(filter(dat_long, outcome == "sens"), by = "study")
    
    metafor::forest(
      dat_sens_plot$yi, vi = dat_sens_plot$vi, slab = dat_sens_plot$study,
      transf = transf.ilogit,
      xlim = c(-1.2, 2.5), 
      ylim = c(-2, k_total + 3), 
      alim = c(0, 1),
      steps = 5,
      rows = all_studies$row_num, header = "Study", 
      main = paste("Sensitivity"), 
      refline = NA, cex = cex)
    
    text(-0.5, k_total + 2, "Instrument", pos = 4, cex = cex, font = 2) 
    text(-0.5, all_studies$row_num, all_studies$instrument_label, 
         pos = 4, cex = cex
    )
    addpoly(res$beta[1], sei = res$se[1],
            row = -1, transf = transf.ilogit, mlab = "", col = "red", cex = cex
    )
    
    # 5. Plot Specificity
    par(mar = c(5, 1, 3, 2))
    dat_spec_plot <- all_studies %>% 
      left_join(filter(dat_long, outcome == "spec"), by = "study")
    
    metafor::forest(
      dat_spec_plot$yi, 
      vi = dat_spec_plot$vi, 
      slab = NA, transf = transf.ilogit, 
      xlim = c(-0.1, 2), 
      ylim = c(-2, k_total + 3), 
      rows = all_studies$row_num,
      alim = c(0, 1),
      steps = 5,
      header = "", main = "Specificity", refline = NA, cex = cex)
    
    addpoly(
      res$beta[2], 
      sei = res$se[2], 
      row = -1, 
      transf = transf.ilogit, 
      mlab = "", col = "red", cex = cex
    )
    
    
    ### Now do pooled ROC
    
    inv_logit <- function(x) { exp(x) / (1 + exp(x)) }

    # Create the study_data data frame for ggplot
    study_data <- data.frame(
      # 1. Back-transform logit specificity to raw specificity, then calculate FPR
      fpr = 1 - inv_logit(dat_spec$yi),
      
      # 2. Back-transform logit sensitivity to raw sensitivity
      sens = inv_logit(dat_sens$yi),
      
      # 3. Pull the strategy names from your original data frame
      # (using the column you cleaned with stringr::word)
      strategy = data[[instrument_col]]
    )
    study_data$weight <- 1 / dat_sens$vi
    
    # 2. Extract the Fixed Effects (Summary Points)
    # (Assuming intercept 1 is logit_FPR and intercept 2 is logit_Se)
    mu_fpr <- res$beta[1] 
    mu_se  <- res$beta[2] 
    
    # 3. Extract Random Effects (Variances and Correlation)
    # 'tau2' contains the variances, 'rho' is the correlation in rma.mv
    sigma_fpr <- sqrt(res$tau2[1]) 
    sigma_se  <- sqrt(res$tau2[2]) 
    rho       <- res$rho
    
    # 4. Generate the Sequence of FPR values for the x-axis
    # We use a sequence from 0.01 to 0.99 to avoid infinity in logit
    fpr_seq <- seq(0.01, 0.99, length.out = 100)
    logit_fpr_seq <- logit(fpr_seq)
    
    # 5. Calculate Expected logit(Sensitivity) using the SROC formula
    logit_se_seq <- mu_se + rho * (sigma_se / sigma_fpr) * (logit_fpr_seq - mu_fpr)
    
    # 6. Back-transform to normal scale (0 to 1)
    sens_seq <- inv_logit(logit_se_seq)
    
    # 7. Create the data frame for ggplot2
    sroc_data <- data.frame(
      fpr = fpr_seq,
      sens = sens_seq
    )
    
    
    mu_fixed <- c(res$beta[1], res$beta[2]) 
    V_fixed  <- vcov(res)[1:2, 1:2] # Variance-covariance matrix of the fixed effects
    
    # 3. Create the Summary Point Data Frame
    summary_point <- data.frame(
      fpr = inv_logit(mu_fixed[1]),
      sens = inv_logit(mu_fixed[2])
    )
    
    # 4. Generate 95% Confidence Ellipse in Logit Space
    # The ellipse() function creates 100 coordinates outlining the region
    ellipse_logit <- ellipse::ellipse(V_fixed, centre = mu_fixed, level = 0.95)
    
    # 5. Back-transform Ellipse Coordinates to Probability Space (0 to 1)
    ellipse_data <- data.frame(
      fpr = inv_logit(ellipse_logit[, 1]),
      sens = inv_logit(ellipse_logit[, 2])
    )

    
    
    plot2 <- 
      ggplot(sroc_data, aes(x = fpr, y = sens)) +
      geom_line(linewidth = 1.2) +
      # Match the visual style of the published plot
      scale_x_continuous(name = "False-Positive Rate", limits = c(0, 1), expand = c(0, 0)) +
      scale_y_continuous(name = "Sensitivity", limits = c(00, 1.0), expand = c(0, 0)) +
      theme_classic() +
      theme(
        legend.position = "bottom",
        legend.title = element_blank(),
        panel.grid.major = element_line(color = "grey95"),
        axis.text = element_text(size = 10),
        axis.title = element_text(size = 12, face = "bold")
      ) +
      geom_path(data = ellipse_data, aes(x = fpr, y = sens), linewidth = 0.8) +
      geom_point(data = study_data, aes(x = fpr, y = sens, size = weight), 
             alpha = 0.5) 
    
    print(plot2)
  }


categories <- c("self_report", "peer_rating", "neuropsycho_tests", "neuroimaging", 
                "clinician_interview", "combination", "biomarker", "EEG")

#| fig-width: 14
#| fig-height: 14
#| 
# Define the categories you want to loop through

for (category in categories) {
  # Dynamically construct column names
  tp <- paste0("true_positive_", category)
  fn <- paste0("false_negative_", category)
  tn <- paste0("true_negative_", category)
  fp <- paste0("false_positive_", category)
  inst <- paste0("test_description_", category)
  
  # Generate the plot
  plot_diagnostic_meta(
    data = d, 
    diag_type_name = category, 
    true_pos = tp, 
    false_neg = fn, 
    true_neg = tn, 
    false_pos = fp, 
    instrument_col = inst)
}

Warning: Removed 92 rows containing missing values or values outside the scale range
(`geom_point()`).

Warning: Removed 129 rows containing missing values or values outside the scale range
(`geom_point()`).

Warning: Removed 109 rows containing missing values or values outside the scale range
(`geom_point()`).

Warning: Removed 120 rows containing missing values or values outside the scale range
(`geom_point()`).

Warning: Removed 127 rows containing missing values or values outside the scale range
(`geom_point()`).

Warning: Removed 120 rows containing missing values or values outside the scale range
(`geom_point()`).

Warning: Removed 124 rows containing missing values or values outside the scale range
(`geom_point()`).

Warning: Removed 119 rows containing missing values or values outside the scale range
(`geom_point()`).